Overview#
openGauss is Huawei's open-source relational database,
forked from PostgreSQL 9.2.4. It speaks the PostgreSQL frontend/backend wire
protocol, so OpenGaussDialect extends PostgresDialect and reuses the
pg driver — connection pooling, $n parameter binding, escaping, and the
bulk of the query builders are inherited unchanged.| Property | Value |
|---|
| Dialect name | opengauss |
| Extends | PostgresDialect (postgres) |
| Driver | pg |
| Cloud SDK | none |
| Default port | 5432 |
| Local runtime | self-hosted; no image in this repo |
The fork point is what makes this dialect more than a rename. Everything
PostgreSQL gained after 9.2 exists in openGauss only if openGauss chose to
backport it, and where it did not — or where it backported a different
spelling — an inherited-verbatim Postgres dialect emits SQL that an openGauss
server refuses. This is the largest set of real divergences of any
compatibility-family dialect in this library.Key facts (from src/dialects/opengauss/index.ts):Dialect name is 'opengauss'. The underlying driver library stays pg.
Default port is 5432, the same as PostgreSQL. No override is needed.
Authentication is the thing most likely to stop you. See the next section
before anything else.
Quick start#
Install#
$ npm install ts-prorm-orm pg
$ pnpm add ts-prorm-orm pg
$ yarn add ts-prorm-orm pg
$ bun add ts-prorm-orm pg
pg is the same driver the postgres dialect uses. pg-copy-streams is optional: only the features that need it load it.Connect#
import { Prorm } from 'ts-prorm-orm';
export const db = new Prorm({
dialect: 'opengauss',
host: 'localhost',
database: 'postgres',
username: 'app',
password: process.env.DB_PASSWORD,
});
import { Prorm } from 'ts-prorm-orm';
export const db = new Prorm({
dialect: 'opengauss',
url: 'opengauss://app:secret@localhost:5432/postgres',
});
A URL is folded into host, port, database, username and password before the dialect is built; any of those passed alongside it wins.Define a model#
import { DataTypes } from 'ts-prorm-orm';
const User = db.define('User', {
id: { type: DataTypes.INTEGER(), primaryKey: true, autoIncrement: true },
email: { type: DataTypes.STRING(255), allowNull: false, unique: true },
name: { type: DataTypes.STRING(100) },
meta: { type: DataTypes.JSONB() },
});
await db.sync(); // CREATE TABLE — see Generated SQL below
meta is JSONB, which the containment query below needs.Query#
import { Op } from 'ts-prorm-orm';
await User.create({ email: 'ada@example.com', name: 'Ada' });
const page = await User.findAll({
where: { name: { [Op.like]: 'A%' } },
order: [['id', 'ASC']],
limit: 20,
offset: 40,
});
Generated SQL#
Each block is the statement the model call above it hands the dialect's query(), captured with the driver stubbed out. The model layer inlines the values.Create table#
db.sync() for the model in the Quick start:CREATE TABLE
Insert, update and delete#
await User.create({ email: 'ada@example.com', name: 'Ada' });
await User.update({ name: 'Grace' }, { where: { email: 'ada@example.com' } });
await User.destroy({ where: { id: 1 } });
Upsert#
await User.upsert({ id: 1, email: 'ada@example.com', name: 'Ada Lovelace' });
upsert() also reads the row by primary key before and after the write, to return the instance.await User.findAll({
where: { name: { [Op.like]: 'A%' } },
order: [['id', 'ASC']],
limit: 20,
offset: 40,
});
JSON#
Containment on the meta column:import { Op } from 'ts-prorm-orm';
await User.findAll({ where: { meta: { [Op.contains]: { plan: 'pro' } } } });
Transaction with a savepoint#
await db.transaction(async (t) => {
await User.create({ email: 'grace@example.com', name: 'Grace' }, { transaction: t });
// A nested transaction is a SAVEPOINT; its failure undoes only its own work.
await db
.transaction(async (sp) => {
await User.destroy({ where: { id: 1 }, transaction: sp });
throw new Error('keep the row');
}, { transaction: t })
.catch(() => {});
});
Table sampling#
sample on the select builder reads a random fraction of the table, for estimates over large tables:const sample = db.getDialectInstance()!.buildSelectQuery({
tableName: 'users',
sample: { method: 'BERNOULLI', percent: 10 },
});
await db.query(sample.sql);
Type mapping for the portable descriptors: STRING → VARCHAR(255), auto-increment INTEGER → SERIAL, BOOLEAN → BOOLEAN, DATE → TIMESTAMP, UUID → UUID, JSON → JSON. The type comparison lines the major engines up side by side.Connection#
⚠️ The pg driver cannot authenticate against a stock openGauss install#
openGauss defaults password_encryption_type to sha256, an
openGauss-proprietary authentication method that neither libpq nor the pure-JS
pg driver implements. A connection through this dialect will fail to
authenticate against a default install, and the error will be about the
authentication method rather than about anything you did.To use the pg driver you must switch the server to md5:1.
ALTER SYSTEM SET password_encryption_type = 0; (or set it in
postgresql.conf) and restart.
2.
Add an md5 line for your host to pg_hba.conf.
3.
Re-set the user's password — the existing one is still stored as a
sha256 hash and will not be re-derived.
The alternative is a driver that implements the sha256 handshake — the
official openGauss-connector-nodejs, or the pg-opengauss fork. This
dialect does not bundle either; it uses pg.This is a deployment prerequisite, not something the dialect can work around.
import { Prorm } from 'ts-prorm-orm';
const db = new Prorm({
dialect: 'opengauss',
host: '127.0.0.1',
port: 5432,
username: 'app',
password: '...',
database: 'postgres',
});
What differs from Postgres#
OpenGaussDialect extends PostgresDialect (postgres) and
overrides 22 of its methods (Object.getOwnPropertyNames(OpenGaussDialect.prototype),
computed by the scratch test); everything else is inherited unchanged.| Area | Overridden |
|---|
| Connection and execution | listen, notify, startTransaction, unlisten |
| DML builders | buildInsertQuery, buildUpsertQuery, formatDefaultValue |
| Types | buildColumnDefinitionSql |
| DDL and schema | addIndex, attachPartition, buildAddPartitionSQL, buildCreatePartitionedTableSQL, buildPartitionSql, createIndex, createPartition, createPartitionedTable, detachPartition |
| Introspection | assertIndexMethod |
| Other | buildDuplicateKeyUpdateClause, buildStorageClause, quoteBound, toOpenGaussPartition |
Divergences from PostgreSQL#
Upsert — openGauss has no ON CONFLICT#
ON CONFLICT arrived in PostgreSQL 9.5, after the fork, and openGauss never
took it. A real server answers:syntax error at or near "CONFLICT"
What openGauss has instead is the MySQL-shaped INSERT ... ON DUPLICATE KEY
UPDATE, which names no conflict target — the server infers it from the table's
unique constraints. buildInsertQuery (with options.upsert) and
buildUpsertQuery emit that form:db.dialect.buildUpsertQuery('kv', { k: 'a', v: 1 });
Two consequences worth calling out:conflictFields is not required. The Postgres base throws without it,
because ON CONFLICT cannot infer its target. openGauss's grammar has nowhere
to put a conflict target at all, so one passed for cross-dialect portability
is accepted and ignored.
VALUES(col), not excluded.col. VALUES(col) is the spelling openGauss
documents for reaching the row that would have been inserted.
updateOnDuplicate restricts the update set as elsewhere:db.dialect.buildUpsertQuery('users', { id: 1, name: 'Ada', email: 'a@x.io' }, {
updateOnDuplicate: ['name'],
});
// ... ON DUPLICATE KEY UPDATE "name" = VALUES("name")
RETURNING after an upsert#
openGauss supports INSERT ... RETURNING for a plain insert — that path is
inherited from the Postgres dialect and works — but not in combination with
ON DUPLICATE KEY UPDATE. Requesting both throws:openGauss does not support RETURNING together with ON DUPLICATE KEY UPDATE
(its only upsert form). Re-fetch the row after the upsert instead.
LISTEN / NOTIFY are not implemented#
openGauss answers LISTEN statement is not yet supported. The Postgres dialect
exposes listen() / unlisten() / notify() and holds a dedicated pg client
for the subscription — so inheriting it would open and keep a connection for a
channel that can never fire. All three throw here instead. Use polling or an
external message queue for change notification.Partitioning is Oracle-shaped, not PostgreSQL 10 declarative#
openGauss did not adopt PostgreSQL 10's declarative partitioning. Every
partition is declared inline in CREATE TABLE, and a range bound is an
exclusive VALUES LESS THAN (...) rather than FOR VALUES FROM (...) TO (...).
Neither statement parses on the other engine.db.dialect.buildCreatePartitionedTableSQL(
'events',
{ id: { type: 'INTEGER' }, ts: { type: 'TIMESTAMP' } },
{
type: 'range',
column: 'ts',
partitions: [
{ name: 'p2024', lessThan: '2025-01-01' },
{ name: 'pmax', lessThan: 'MAXVALUE' },
],
},
);
List partitions carry a bare value set (VALUES ('de', 'fr'), or
VALUES (DEFAULT) for the catch-all); hash partitions carry only names, since
openGauss derives the buckets from the partition count.Because there is no CREATE TABLE ... PARTITION OF, a range or list partitioned
table must declare its partitions up front — asking for one with an empty
partition list throws rather than emitting a table that cannot be created.createPartitionedTable() accepts the same { partitionBy, partitions } option
shape as the Postgres base, translating a shared PartitionDefinition's to
bound into VALUES LESS THAN. A range partition's from has no openGauss
counterpart: partitions are contiguous, and each one's lower bound is the
previous one's upper bound.Later partitions are added with ALTER TABLE ... ADD PARTITION, and only at the
end of the range:db.dialect.buildAddPartitionSQL({ parentTable: 'events', name: 'p2026', bound: { from: 0, to: 100 } });
attachPartition() and detachPartition() throw. openGauss has neither
ATTACH PARTITION nor DETACH PARTITION; the nearest operations are
ADD PARTITION for a genuinely new partition and EXCHANGE PARTITION to swap
in an identically-shaped table. Neither preserves a detached partition as a
standalone table the way DETACH does, so they are not silently remapped.Index methods — no BRIN, no SP-GiST#
| Method | Where it applies |
|---|
btree | Row-store (Astore), the default |
ubtree | Row-store (Ustore engine) |
hash | Row-store |
gin | Row-store and column-store |
gist | Row-store |
psort | Column-store, the default there |
PostgreSQL's brin and spgist are not available. Passing either to
addIndex() or createIndex() throws up front, naming the dialect, rather than
reaching the server as CREATE INDEX ... USING brin (...) and failing with an
access-method error that mentions nobody.Column-store tables#
WITH (ORIENTATION = COLUMN) creates a column-oriented (analytics) table — an
openGauss storage option with no PostgreSQL equivalent.db.dialect.buildStorageClause({ orientation: 'COLUMN', compression: 'HIGH' });
// " WITH (ORIENTATION = COLUMN, COMPRESSION = HIGH)"
A table that requests neither produces an empty clause, so ordinary row-store
DDL stays byte-identical to what the Postgres dialect emits.Accepted, but not what it says#
These emit valid SQL and are therefore not rejected — but the semantics differ,
so relying on the PostgreSQL meaning is a bug.SERIALIZABLE is really REPEATABLE READ#
openGauss accepts the keyword and runs the transaction at REPEATABLE READ. There
is no serializable-snapshot isolation, so there is no protection against write
skew. startTransaction() warns rather than throwing, because the statement
does execute — failing the transaction outright would be worse than saying so:openGauss maps SERIALIZABLE to REPEATABLE READ: it has no serializable-snapshot
isolation, so this transaction is not protected against write skew.
Other inherited-but-different behaviour#
SERIAL on a temporary table is rejected by openGauss
(It's not supported to create serial column on temporary table). SERIAL
on an ordinary table works exactly as in PostgreSQL.
Binary COPY differs subtly from PostgreSQL's and can fail with
unexpected EOF against PostgreSQL-native binary COPY tooling. Text-format
COPY is unaffected.
nextval() runs only on the primary node, never on a standby — relevant
if you route reads to a standby in an HA topology.
Inherited PostgreSQL behavior#
Everything not listed above is inherited from PostgresDialect and behaves as
documented for the PostgreSQL dialect — double-quoted
identifiers, $n placeholders, SELECT/UPDATE/DELETE builders, JSON/JSONB
helpers, LIMIT/OFFSET, FOR UPDATE locking, and the type mapping.openGauss does support, and so these are left alone: jsonb, uuid, arrays,
interval, range types, TABLESAMPLE, CREATE INDEX CONCURRENTLY (but not on
column-store, temporary or partitioned tables, and not inside a transaction
block), event triggers, CREATE PROCEDURE, anonymous DO blocks, and
NOWAIT row locking.Caveats / not-yet-verified#
SQL generation is verified; live execution is not. Every SQL snippet on
this page is the exact output of a builder asserted in
tests/dialects/opengauss.test.ts, which runs without any database
connection. None of it has been executed against a live openGauss server in
this repository — and given the sha256 authentication prerequisite above, a
connection is a deployment exercise in its own right.
SKIP LOCKED is unconfirmed. NOWAIT is documented; SKIP LOCKED was
not found in openGauss's locking documentation either way. The inherited lock
builder will emit it if asked. Verify before relying on it.
Several PostgreSQL features could not be confirmed either way because the
official documentation pages did not render fetchable body text: the FILTER
aggregate clause, GROUPING SETS/CUBE/ROLLUP, LATERAL joins, INCLUDE
covering indexes, pg_trgm, and TRUNCATE ... RESTART IDENTITY. All are
emitted by inherited builders and none is rejected here. Test them against
your server rather than assuming the PostgreSQL answer.
Reserved words have not been diffed. No openGauss-vs-PostgreSQL keyword
diff was available. All identifiers are double-quoted by the inherited
escaping, which covers the common case; run
SELECT * FROM pg_get_keywords() against both engines if you need precision.
Introspection helpers inherited from Postgres query pg_catalog /
information_schema and have not been checked against openGauss's catalogs.
GENERATED ALWAYS AS IDENTITY and stored generated columns are reported as
supported but were not confirmed from primary documentation.
Capabilities#
Computed from the code — capabilitiesOf() (src/transaction-support.ts),
whereSupportOf() (src/query-builders/expression-render.ts),
sqlObjectCapabilitiesOf() (src/sql-objects/capabilities.ts) and the
builders — by the scratch test described in the
capability matrix.| Capability | opengauss |
|---|
| Transactions atomic | yes |
| Savepoints | yes |
| Isolation | up to SERIALIZABLE |
| Transaction modes | read-only yes; deferrable yes; constraint deferral yes |
| RETURNING | RETURNING on INSERT |
| Upsert | INSERT … ON DUPLICATE KEY UPDATE |
| JSON column | JSON (JSONB → JSONB) |
| Vector column | VECTOR(3) |
| Array column | INTEGER[] |
| WHERE builder | 62 operator keys; $and/$or/$not; $exists: yes; $in subquery: yes; expression operands: yes |
| SQL objects | family unknown: triggers no; procedures no; functions no; materialized views no; sequences no; enums check; RLS no |
| Streaming | queryStream() present; LISTEN present; createExtension() present |
Querying — how a FindOptions becomes SQL Modified at 2026-10-04 00:40:16